In a torque converter provided with a lock-up clutch, control of a front-rear differential pressure (lock-up differential pressure) of the lock-up clutch engages and releases the lock-up clutch. To shift the torque converter from a converter state to a lock-up state, the lock-up differential pressure gradually increases from a predetermined initial differential pressure. The torque converter shifts from the converter state to the lock-up state via a slip state. In the converter state the lock-up clutch is released, in the slip state the lock-up clutch slips, and in the lock-up state, the lock-up clutch is engaged.
In this lock-up clutch control, the real lock-up clutch differential pressure has scatter due to individual differences and time-dependent variations of the torque converter. JP2000-27986 published by the Japan Patent Office in 2000 discloses a prior art technique wherein learning control of the differential pressure is performed in order to correct the deviation of engagement timing due to this scatter.